In [1]:
%reload_ext autotime
import pandas as pd
import requests
from pprint import pprint
import json
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
from tqdm.auto import tqdm

pd.options.plotting.backend = "plotly"
pd.set_option("display.max_columns", None)
pd.set_option("display.max_colwidth", 100)

✔️ 4.57 s (2024-12-16T12:12:26/2024-12-16T12:12:30)
2024-12-16 12:12:29.284009: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-12-16 12:12:29.296515: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-12-16 12:12:29.313946: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-12-16 12:12:29.319666: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-12-16 12:12:29.334112: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-12-16 12:12:30.150251: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
In [2]:
df = pd.read_csv("results.csv").drop_duplicates(subset="panoid")
df

✔️ 25.7 ms (2024-12-16T12:12:30/2024-12-16T12:12:30)
Out[2]:
Index pid n time anxiousness latitude longitude geometry panoid panolat panolon panodate panothirdparty source uploader
0 0 P20001 1 2023-04-25T02:51:42Z 0 -36.924795 174.738044 POINT (174.7380435 -36.92479483) IvrcS0W1RlFAlnci-p39XA -36.924665 174.737914 2012-04 False launch NaN
10 10 P20001 11 2023-04-24T00:42:25Z 0 -36.924837 174.737948 POINT (174.7379477 -36.92483659) QEpZV7bnO2mBfp0weMUKEg -36.924730 174.737826 2012-04 False launch NaN
13 13 P20006 1 2023-06-03T02:45:55Z 3 -36.892203 174.740125 POINT (174.7401253 -36.89220256) omb98QNjTPWi0uUfMsmYeg -36.892621 174.739961 2024-05 False launch NaN
14 15 P20009 2 2023-05-17T04:54:48Z 3 -36.923191 174.748620 POINT (174.7486203 -36.92319093) E7B5AV3DQ1rYWDClVRo8Zg -36.923194 174.748831 2024-05 False launch NaN
17 19 P20009 6 2023-05-19T22:28:51Z 1 -36.923260 174.748655 POINT (174.748655 -36.92325959) KCTcsxYCIm41XdzkYEYUQw -36.923286 174.748840 2024-05 False launch NaN
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1314 1421 P20693 2 2024-05-02T03:43:23Z 3 -36.897778 174.721580 POINT (174.7215796 -36.89777786) Uzuqd6oSo-EjCVuRP2Os0Q -36.897742 174.721877 2022-06 False launch NaN
1317 1425 P20693 6 2024-05-05T03:00:22Z 2 -36.969426 174.790602 POINT (174.7906024 -36.96942642) 4OskePS4Ilz12JhsP-1ujg -36.969164 174.790848 2022-08 False launch NaN
1318 1426 P20721 1 2024-05-05T02:00:52Z 1 -36.893455 174.728262 POINT (174.728262 -36.89345532) CfRtPfDMNhfXHTNvMwnYRw -36.893394 174.728062 2024-06 False launch NaN
1320 1428 P20721 3 2024-05-05T23:06:27Z 2 -36.845252 174.759951 POINT (174.7599508 -36.8452515) AF1QipN2FD2eYEmK8bRpEgoM7fFl5-nUstwWujnRj0gv -36.845292 174.759939 2022-06-24 True photos:street_view_publish_api Mint Design
1321 1429 P20721 4 2024-05-06T07:04:57Z 0 -36.845165 174.759885 POINT (174.7598849 -36.84516487) AF1QipNj6yheGtCvR6Gk2Svq_lG_fuaGPjehPV8kouy8 -36.845177 174.759792 2022-06-24 True photos:street_view_publish_api Mint Design

592 rows × 15 columns

In [3]:
# Loading this model needs about 22.69GB of GPU memory
model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"

model = MllamaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)

✔️ 12.6 s (2024-12-16T12:12:31/2024-12-16T12:12:43)
The model weights are not tied. Please use the `tie_weights` method before using the `infer_auto_device` function.

Loading checkpoint shards:   0%|          | 0/5 [00:00<?, ?it/s]
In [13]:
prompt = """
    This image is a panorama from Google Street View.
    From the image, extract the following information, in JSON format:

    green: The percentage of the image that is green space (e.g., parks, gardens, trees, grass). A number between 0 and 100.
    environment: The general classification of the environment based on the visible surroundings. Choose the closest matching category from the following: "low density residential", "medium density residential", "indoor", "park", "commercial", "shops", "cafes", "supermarket" or suggest a custom classification as a string.
    active_transport: Indicate if an active transport corridor is visible (e.g., bike lane, walking path). Return true or false.
    quality: A subjective assessment of the area's upkeep, where 0 represents "run down" and 100 represents "well maintained." A number between 0 and 100.
    water: If streams, ponds, rivers, or the ocean are visible, estimate the distance to the nearest body of water in meters. If no water is present, return 0.
    obscured: The proportion of the view obscured by buildings (i.e., how much of the total line of sight is blocked by buildings in close proximity). A number between 0 and 100.
    people: The total number of people visible in the image. A whole number.
    cars: The total number of cars visible in the image. A whole number.
    bikes: The total number of bikes visible in the image. A whole number.
    dwellings: The total number of dwellings (houses, flats, or apartments) visible in the image. A whole number.
    shops: The total number of shops visible in the image. A whole number.
    offices: The total number of offices visible in the image. A whole number.

    Do not include comments in your JSON response. Only respond with the JSON object. Make sure the JSON is valid.
"""
for row in tqdm(df.sample(10).itertuples(index=False)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.jpg")
    display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": prompt},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        print("Output:")
        try:
            result = json.loads(result)
            pprint(result)
            print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

✔️ 40 s (2024-12-16T14:34:12/2024-12-16T14:34:52)
0it [00:00, ?it/s]
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 1,
 'cars': 5,
 'dwellings': 5,
 'environment': 'commercial',
 'green': 40,
 'obscured': 20,
 'offices': 0,
 'people': 10,
 'quality': 70,
 'shops': 3,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 0,
 'cars': 1,
 'dwellings': 6,
 'environment': 'low density residential',
 'green': 50,
 'obscured': 30,
 'offices': 0,
 'people': 1,
 'quality': 60,
 'shops': 1,
 'water': 200}
No description has been provided for this image 
Output:
{'active_transport': True,
 'bikes': 0,
 'cars': 1,
 'dwellings': 4,
 'environment': 'low density residential',
 'green': 58,
 'obscured': 30,
 'offices': 0,
 'people': 1,
 'quality': 80,
 'shops': 0,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 0,
 'cars': 5,
 'dwellings': 3,
 'environment': 'medium density residential',
 'green': 20,
 'obscured': 30,
 'offices': 0,
 'people': 0,
 'quality': 60,
 'shops': 1,
 'water': 200}
No description has been provided for this image 
Output:
{'active_transport': True,
 'bikes': 2,
 'cars': 20,
 'dwellings': 15,
 'environment': 'medium density residential',
 'green': 20,
 'obscured': 40,
 'offices': 3,
 'people': 10,
 'quality': 80,
 'shops': 5,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 0,
 'cars': 2,
 'dwellings': 10,
 'environment': 'low density residential',
 'green': 30,
 'obscured': 20,
 'offices': 0,
 'people': 0,
 'quality': 50,
 'shops': 0,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 0,
 'cars': 1,
 'dwellings': 3,
 'environment': 'low density residential',
 'green': 40,
 'obscured': 10,
 'offices': 0,
 'people': 0,
 'quality': 70,
 'shops': 1,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': True,
 'bikes': 0,
 'cars': 16,
 'dwellings': 2,
 'environment': 'commercial',
 'green': 34,
 'obscured': 44,
 'offices': 0,
 'people': 2,
 'quality': 75,
 'shops': 1,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': False,
 'bikes': 0,
 'cars': 4,
 'dwellings': 11,
 'environment': 'low density residential',
 'green': 60,
 'obscured': 20,
 'offices': 0,
 'people': 0,
 'quality': 70,
 'shops': 0,
 'water': 0}
No description has been provided for this image 
Output:
{'active_transport': True,
 'bikes': 2,
 'cars': 15,
 'dwellings': 10,
 'environment': 'low density residential',
 'green': 50,
 'obscured': 20,
 'offices': 0,
 'people': 0,
 'quality': 80,
 'shops': 3,
 'water': 0}
In [14]:
results = []
for row in tqdm(df.itertuples(index=False), total=len(df)):
    panoid = row.panoid
    image = Image.open(f"panoramas/{panoid}.jpg")
    #display(image)
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "text", "text": prompt},
                {"type": "image"},
            ]
        }
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(
        image,
        input_text,
        add_special_tokens=False,
        return_tensors="pt"
    ).to(model.device)

    for retry in range(3):
        output = model.generate(**inputs, max_new_tokens=5000)
        result = processor.decode(output[0])
        result = result[result.rindex("<|end_header_id|>") + len("<|end_header_id|>"):].strip().replace("<|eot_id|>", "")
        #print("Output:")
        try:
            result = json.loads(result)
            #pprint(result)
            row = row._asdict()
            row.update(result)
            results.append(row)
            #print("\n")
            break
        except json.JSONDecodeError:
            print(f"Unable to parse: {result}")

results = pd.DataFrame(results)
results.to_csv("LLM_results.csv", index=False)
results

✔️ 37 min 44 s (2024-12-16T14:35:11/2024-12-16T15:12:56)
  0%|          | 0/592 [00:00<?, ?it/s]
Out[14]:
Index pid n time anxiousness latitude longitude geometry panoid panolat panolon panodate panothirdparty source uploader green environment active_transport quality water obscured people cars bikes dwellings shops offices
0 0 P20001 1 2023-04-25T02:51:42Z 0 -36.924795 174.738044 POINT (174.7380435 -36.92479483) IvrcS0W1RlFAlnci-p39XA -36.924665 174.737914 2012-04 False launch NaN 25 low density residential False 60 0 0 0 2 0 7 2 0
1 10 P20001 11 2023-04-24T00:42:25Z 0 -36.924837 174.737948 POINT (174.7379477 -36.92483659) QEpZV7bnO2mBfp0weMUKEg -36.924730 174.737826 2012-04 False launch NaN 65 low density residential True 80 0 20 0 1 0 8 0 0
2 13 P20006 1 2023-06-03T02:45:55Z 3 -36.892203 174.740125 POINT (174.7401253 -36.89220256) omb98QNjTPWi0uUfMsmYeg -36.892621 174.739961 2024-05 False launch NaN 60 medium density residential False 70 100 30 0 1 0 15 1 0
3 15 P20009 2 2023-05-17T04:54:48Z 3 -36.923191 174.748620 POINT (174.7486203 -36.92319093) E7B5AV3DQ1rYWDClVRo8Zg -36.923194 174.748831 2024-05 False launch NaN 44 low density residential False 70 0 25 0 7 0 17 2 0
4 19 P20009 6 2023-05-19T22:28:51Z 1 -36.923260 174.748655 POINT (174.748655 -36.92325959) KCTcsxYCIm41XdzkYEYUQw -36.923286 174.748840 2024-05 False launch NaN 40 low density residential True 70 120 30 5 4 1 10 2 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
587 1421 P20693 2 2024-05-02T03:43:23Z 3 -36.897778 174.721580 POINT (174.7215796 -36.89777786) Uzuqd6oSo-EjCVuRP2Os0Q -36.897742 174.721877 2022-06 False launch NaN 25 low density residential False 60 0 20 1 2 0 8 1 0
588 1425 P20693 6 2024-05-05T03:00:22Z 2 -36.969426 174.790602 POINT (174.7906024 -36.96942642) 4OskePS4Ilz12JhsP-1ujg -36.969164 174.790848 2022-08 False launch NaN 30 low density residential False 60 0 40 0 2 0 7 0 0
589 1426 P20721 1 2024-05-05T02:00:52Z 1 -36.893455 174.728262 POINT (174.728262 -36.89345532) CfRtPfDMNhfXHTNvMwnYRw -36.893394 174.728062 2024-06 False launch NaN 65 low density residential False 60 200 20 1 2 0 3 1 0
590 1428 P20721 3 2024-05-05T23:06:27Z 2 -36.845252 174.759951 POINT (174.7599508 -36.8452515) AF1QipN2FD2eYEmK8bRpEgoM7fFl5-nUstwWujnRj0gv -36.845292 174.759939 2022-06-24 True photos:street_view_publish_api Mint Design 0 commercial False 80 0 70 0 6 0 4 4 2
591 1429 P20721 4 2024-05-06T07:04:57Z 0 -36.845165 174.759885 POINT (174.7598849 -36.84516487) AF1QipNj6yheGtCvR6Gk2Svq_lG_fuaGPjehPV8kouy8 -36.845177 174.759792 2022-06-24 True photos:street_view_publish_api Mint Design 0 low density residential False 80 0 0 0 2 0 2 1 0

592 rows × 27 columns